feat(wkb): add MultiPolygon support to WKB parser

[wietzesuijker]

Adds WKBType.MultiPolygon case to parseWkb(). Based on the WKB parser from the kyle/wkb branch, adding one more nesting level for geom-to-polygon offsets using polygonIndices from loaders.gl's BinaryPolygonGeometry.

MultiPolygon is a core OGC Simple Features type. The types (MultiPolygonData, WKBType.MultiPolygon) and type guards already exist in this library, only the parseWkb() case was missing.

Changes

File	Change
src/io/wkb.ts	repackMultiPolygons() + inferMultiPolygonCapacity()
src/io/wkb.test.ts	4 tests: two-polygon, single-polygon, with-hole, multi-feature batch
src/type.ts, src/data.ts	LargeWKB / LargeWKBData type aliases

Tests verify offset arrays at every nesting level and spot-check coordinate values. WKB hex generated via Shapely. All 8 tests pass (4 new + 4 existing).

Note

Based on kyle/wkb. stac-map depends on smohiudd/geoarrow-js (a fork of that branch) because it was never merged upstream.

AI (Claude) supported my development of this PR.

[wietzesuijker]

pushed 30c751d to make the linter happy

[kylebarron]

I think the WKB parser really needs a wholesale re-design. Or, perhaps the design is fine, but there are a lot of considerations

Perhaps you'd be able to point Claude at this folder in geoarrow-rs and enough test cases and it would be able to create a WKB -> GeoArrow parser.. but I'm skeptical that a claude-created version would be possible to maintain without a lot of additional input effort.

I do think the Rust implementation there is quite stable and cleanly-designed, but it's reasonably verbose. You can read, say, the wkb parsing/encoding from cast.

I think you'd need something like https://github.com/developmentseed/geotiff-test-data: a considerable corpus of test data for WKB - GeoArrow interop to have confidence in the code.

That said I don't really have the time to review big changes here at the moment.

[kylebarron]

I definitely wouldn't want to depend on loaders.gl from geoarrow-js.

[kylebarron]

You'd want a type for LargeWKB too

[wietzesuijker]

Thanks @kylebarron. Agreed on the bigger picture. This PR is just adding the missing MultiPolygon case to the existing parser (one function + tests), because I want to use that elsewhere. Happy to convert to draft if you'd rather revisit the whole thing at once.